Professional Documents
Culture Documents
1990,29,931-942 93 1
Catalytic cracking converts heavy petroleum fractions into gasoline, distillates, and light olefins.
It revolutionized refining over 50 years ago and has been evolving continuously since. Catalytic
cracking was commercialized in fixed, moving, and fluid-bed reactors (fluid catalytic cracking, FCC,
has created fine powder fluidization). The design aspects of this important process, in three major
reactor types, constitute a fascinating lesson in reaction engineering. We examined the properties
of catalytic cracking, its chemistry, kinetics, and thermodynamics, and their influence on design.
Of the hundreds of complex reactions, catalytic and thermal, it is the conversion of hydrocarbon
t o coke and gas that dominates reactor and regenerator designs. The effects of hydrodynamics on
cracking are examined through the concept of contact time distribution, and the discussion of the
various reactor and regenerator types concludes with a summary of process control.
-
The first, and one of the most important, part of a re- such as
actor design is the understanding of process chemistry.
The complex processes that occur on a catalytic surface C10H20 2C5H10 (3)
-
during the cracking process are poorly understood. How- Olefins can also polymerize
ever, for a reaction engineer, understanding the chemistry
has a very specific meaning, and Hugh Hulburt was one n-C5H10 C5nH10n (4)
of the first to clearly define it. What is needed is an
understanding of the overall reactions, their thermody-
namics, and the way overall reaction rates are affected by -
and can also undergo hydrogen-transfer reactions such as
C1oH2o CTH8 + C& + 2H2 (5)
t
Yield Vol% Yield. Wt% 5
Gibbs 0
€4 70 80 90 1W €4 70 80 90 1W
Con"err!on,VIllS Canversion. VoI%
1 Amorphous Catalyst
__
60 I I I 2 Zeolhe Catalyst
0 200 400 600 800 Figure 2. Gasoline and coke yields vs conversion for zeolite (on
Temperature, "C silica-alumina support matrix) and silica-alumina (amorphous)
Figure 1. Effect of temperature on Gibbs free energy of cracking catalysts (based on FCC kinetic model (Jacob et al., 1976)).
for a model compound.
For coke formation, this results in the Voorhies (1945)
Olefinic polymers and other high molecular weight equation:
compounds with low hydrogen content, plus low molecular
weight gases such as CHI and H2, are formed. Some of C, = At," (10)
these higher molecular weight compounds form "coke". The exponent n = 0.5 was related by Voorhies to a diffu-
Overall, one can write
C,H,, -
cake f gas (7)
It is this reaction that dominates the design of catalytic
sion mechanism of coke precursors through a coke layer.
In differential form, eq 10 is
dC,/dt, = nAt,"-l (11)
cracking reactors.
One important question in reactor design is the tem- The average catalyst residence time, t , is simply related
perature range that minimizes thermodynamic constraints to the space velocity and the catalyst to oil (C/O) ratio by
on conversion or the range of the desirable reaction AG,
which is significantly negative (preferably less than -5
kcal/mol). There is a large number of individual com-
pounds involved in catalytic cracking, but some conclusions Combining the principles of Voorhies (1945) and Blanding
can be drawn from model compounds. Cracking reactions (1953), Weekman and Nace (1970) represented cracking
are endothermic, but as the AH is significantly larger than as an apparent second-order reaction:
the AG, AG decreases with increasing cracking tempera-
ture. For one model compound (Figure l),AG is -5.0 at dX/dt, = k,Q,(100 - X)' (13)
300 "C. Generally, the lower the molecular weight of a where X is conversion, typically defined as 100 - volu-
compound, the higher is the temperature required to crack metric yields of products boiling higher than gasoline (220
it. Thermodynamics thus restricts the lowest temperature "C). Note that this "conversion" is really a selectivity.
for cracking for any catalyst to approximately 300-400 "C. Cracking is a high conversion process in the sense that
3. Kinetics of the Cracking Reactor almost every molecule in the feed undergoes some change,
but conversion as used here is typically 70-80 vol %.
Cracking is a complex process since there are thousands Q, is a catalyst decay function. Two forms of Q, can
of individual compounds in the feed and in the products. represent cracking data: a first-order decay (exponential)
Fundamental kinetic modeling, following all compounds, and an nth power decay similar to eq 9. In addition, it is
has not been performed as of yet and may be unnecessary. assumed that the cracking activity declines in proportion
Many of the species present can be lumped into groups. to the coking activity. Furthermore, Weekman and Nace
The more lumps, the more difficult is the mathematical assumed that the decay function for gasoline cracking (k2)
treatment and the more accurate is the description of the is of the same shape as eq 13 with the same but with a
system that is needed (including feed analysis). Weekman different rate constant. Many more complex mechanisms
(1969) has shown that, for reactor design purposes, a simple have been proposed, but the mechanism of eq 8 is suffi-
three-lump model is sufficient, as it contains the salient cient to illustrate basic principles. One of the most useful
feed -
features of the system important for the designer:
kt
gasoline (8)
lumped models published, and the one we use here too,
is the one published by Jacob et al. (1976).
A ratio of coke yield to normalized conversion (or
\ / . "crackability") has been used for many years to analyze
coke + gas cracking data:
Reactions 1 and 2 are pseudo second order (second order k , = coke yield/crackability = C(lO0 - X ) / X (14)
in mole fraction but not in partial pressure), whereas re-
action 3 is pseudo zero order in mole fraction and therefore The typical gasoline yield over zeolite and Si/Al catalysts
independent of conversion. Reaction rates depend on the in gas oil cracking is shown in Figure 2A as a function of
catalyst used, operating conditions, and feed properties. conversion. The gasoline yield increases with conversion
All three reaction rates are affected by catalyst deactiva- up to a maximum and then decreases as reaction 2 in eq
tion, which has a permanent component and a regenerable 8 predominates, and gasoline cracks to lighter products and
part (due to coking). One can write for atl three reactions to coke (coke yield is shown in Figure 2B).
kj = k 10. t cn-1 (9)
where Iz,, is the catalyst activity for the regenerated cat- 4. Cracking Catalyst Deactivation by Coke
alyst, n varies from 0.4 to 0.6 for silica-alumina catalysts Presently known cracking catalysts deactivate quickly
(n is lower for zeolites), and t , is the average residence time by coke deposition. Some coke is formed by polymeriza-
of catalyst in the hydrocarbon environment. tion and condensation of primary cracking products. Coke
Ind. Eng. Chem. Res., Vol. 29, No. 6 , 1990 933
I_\ \ i
A. HIC Ratio as Function of 1.Hexane 6 : HIC Ratio in Combusted or 530
Conversion (Adapted from Wolciechowski and Stripped Coke
Coma, 1986)
520
Temp. ‘C
Required
Ratio
HIC
Aiomic HIC tor Constant
Atomic Conversion
Ratio 510
0 500
1 0 z a 3 0 4 0 ~ 0 5 10 15 20
1-Hexane Convenlon pi%) Stripping or Combustion Time
Coke Content on Catalyst, %
Figure 3. H/C ratio in coke. Figure 5. Effect of coke content on activity of Pt reforming catalyst.
10000 r
2
Cracking
1000 I
Catalyst 100 - Coke,
% on 1
Activity
Catalyst
Amorphous
Silica-Alumina
Silica-Alumina
1 -
0 1 2 0
Coke Content on Catalyst, % 0 10 20 30 40 50
Time on Streams, Sec.
Figure 4. Cracking catalyst activity vs coke content. Figure 6. Coke deposition on zeolite and silica-alumina catalysts
as a function of time.
is not a clearly defined compound but a complex mixture
of hydrocarbons. Coke also contains sulfur and nitrogen Coking occurs rapidly, and the coking rate is a function
compounds, which are the source of SO, and NO, emis- of catalyst properties, feed composition, temperature, and
sions in the regenerator flue gas. While it is often assumed pressure. For a given temperature, pressure, and feed
that coke is mostly carbon, in the form of combined aro- composition, the coke level is simply a function of time.
matic rings, other compounds are also present. A con- A typical coke-time curve is shown in Figure 6 for silica-
siderable fraction of the coke is easily strippable by steam, alumina and zeolite catalyst. Comparison of Figure 6 to
or other gases, at typical reaction temperatures (540 OC). Figure 4 shows that zeolite catalyst loses over 90% of its
At higher temperatures, gasification by steam may also initial activity after 1s, whereas silica-alumina loses this
occur. The history of catalytic cracking has been domi- amount of activity in over 1 min. Coking deactivates a
nated by the competition between cracking and coking. cracking catalyst within a much shorter time period than
Some fraction of the coke deposited on spent cracking in almost any other catalytic process. The catalyst
catalyst comes directly from unconverted feed molecules. therefore requires frequent regeneration, and it is this
These can include complex structures such as asphaltenes. frequent regeneration that dominates the design. There
A large fraction of the Conradson carbon residue (CCR) is another aspect of catalyst regeneration that dominates
in the feed is incorporated into coke. The typical H/C design consideration. As approximately 5 wt 9% of the feed
atomic ratio of coke for one model compound (1-hexene) is converted to coke, it is important that the heat evolved
is shown in Figure 3A for a common cracking catalyst, a in combustion be recovered.
Y zeolite. The initial coke formed has a H/C ratio of 2, The other mechanism of catalyst deactivation is
a ratio close to the H/C ratio in the feed and the major “permanent” deactivation by dealumination and metal
product-gasoline and olefinic hydrocarbons. As conver- poisoning. This component of deactivation cannot be
sion increases, the coke becomes more aromatic. The high recovered in the cracking reactor-regenerator system.
H/C %eke" is the first to be removed by stripping or by Dealumination is caused by steaming in the regenerator,
combustion (Figure 3B). The more “graphitic”, which is proportional to steam partial pressure, regener-
‘condensed”, or “residual” carbon is removed last, at a ation temperatures (in an Arrhenius form), and catalyst
slower rate. High-efficiency combustors can reduce re- activity to a power larger than one. Metal poisoning by
sidual carbon on regenerated catalysts down to less than sodium, nickel, and vanadium is more important as heavier
0.05 w t %. feeds (such as residuum) are processed. Metals buildup
The cracking activity decreases with the coke content on the catalyst increases coke and gas yields and accel-
of the catalyst. A typical plot for pure zeolite and an erates catalyst deactivation. Catalyst makeup require-
amorphous silica-alumina catalyst is given in Figure 4. ments in catalytic cracking are determined by the need to
During burnoff, the coke is redistributed, and residual maintain a desired activity level to compensate for these
coke, while reducing activity, has a lower impact as com- deactivation mechanisms.
pared to fresh coke. For comparison, the activity of a
reforming catalyst as a function of coke content is shown 5. Thermal Cracking
in Figure 5. The effect of coke on activity is much smaller Another important competing reaction is thermal
in reforming, and the catalyst can have a much higher coke cracking. Thermal cracking was first used for the pro-
content before most activity is lost (20% versus 1% in duction of gasoline from fuel oil 80 years ago and is still
catalytic cracking). used today in resid cracking, visbreaking, and coking op-
934 Ind. Eng. Chem. Res., Vol. 29, No. 6, 1990
'"h
thermal cracking can be increased by increasing the ac-
tivity of the catalyst. If the ratio of thermal to catalytic
cracking is fixed, then a given catalyst activity per unit 25
volume results in a maximum permissible temperature and
ResIdence
a maximum permissible residence time. Higher temper- The,
atures reduce the ratio of catalytic coking rate to catalytic Minutes
cracking rate and also improve some product properties,
such as octane.
(3) Uniform Contact Time Distribution. The contact
time distribution (Shinnar et al., 1973; Shinnar, 1986) is
600 650 700 750
a dimensionless measure of how conversion and product Temperature,'
C
distribution in the reactor deviate from plug flow. Uniform
Figure 11. Effect of temperature on coke burning rate.
contact time improves selectivity.
The concept of a residence time distribution has had a the average gas residence time, or 5-25 s.) Therefore, if
strong impact on reaction engineering over the past 30 thermal reactions affect molecules absorbed on the Catalyst
years. But it only applies rigorously to homogeneous re- surface, then there will be a significant difference between
actors, as it assumes that all molecules stay in the reactor an entrained bed and a stationary catalyst type of reactor
the same amount of time regardless of their state. In (such as many laboratory-scale FCC reactors (Sapre and
heterogeneous systems, molecules are absorbed at a sur- Leib, 1990)).
face, and adsorption and desorption are selective: the real The flow pattern can deviate considerably from plug
sojourn times of molecules can be different for different flow, if catalyst densities vary across the cross section or
species. Furthermore, it is not residence time in the reador if the catalyst phase has significant backmixing due to
that counts but the time in contact with a catalytic surface. clustering. \k(t)an be strongly different from plug flow,
For a first-order reaction, one can define a contact time which can lead to yield penalties. A uniform \k(t)also
distribution rigorously. If we assume a first-order reaction requires that there would be no diffusional resistances
network with rate matrix Ri, then the outlet composition inside the catalyst particle. This limits the particle size.
in a plug flow reactor is
9. Kinetics of Coke Combustion
+
Cj = Ajo CAjnCXnr (15) Catalytic cracking yields large amounts of coke (4-8 w t
where Aj,, are constants determined by the feed compo- % of feed). Hydrogen-rich compounds either volatilize at
sition, A,, are the eigenvalues of the rate matrix, and 7 is the temperatures of a typical regenerator (650-800 "C) or
a time constant related to space velocity. For any iso- crack to combustible volatiles and coke. The time required
thermal catalytic reactor with arbitrary flows, one can show for catalyst regeneration is dominated by the rate of 'slow
in a similar fashion to a homogeneous reactor that outlet coke" combustion. The activation energy of the combus-
composition is tion is approximately 35 kcal/mol. A typical curve of
reaction rate versus temperature is given in Figure 10.
+
Cj = Ajo CAjn@(Ant) (16) The time required for regenerating a catalyst with 1%
The funciton $ ( A n t ) is the Laplace transform of the coke (with 1%oxygen breakthrough) is given in Figure 11
function W). The contact time distribution is only correct as a function of temperature. This time scale is important
for the specific pseudo-first-order reaction matrix for which as it fixes catalyst inventory and regenerator size.
it was measured. It is a quantitative measure of how the Reducing the size of the regenerator and the catalyst
flow deviates from plug flow. inventory are important for two reasons: FCC regenerators
As molecules are absorbed on a catalyst surface (or in- are large and their cost is a significant fraction of total unit
side zeolite crystals), the actual sojourn time of a molecule cost. Furthermore, minimizing catalyst inventory has
in the reactor can be much larger than the residence time significant advantages not only in reducing operating costa
in the gas phase. Depending on absorption, the difference but also in allowing faster adjustment of the inventory to
can be an order of magnitude. There is virtually no data changing needs of feed or product composition.
in the literature to account for this effect, which is very FCC units operate in "heat balance", and hot catalyst
important in a fixed or fixed-fluid bed reactor. On the from the regenerator supplies part of the heat of cracking.
other hand, in an entrained-bed reactor such as an FCC The FCC unit heat balance depends on catalyst activity,
riser, the sojourn time of any molecule is limited to the feed properties, feed preheat, and reactor temperature.
residence time of the catalyst particles. (Typically, the The heat balance also depends on the desired C02/C0
average solids residence time in the FCC reactor is 3 times ratio in the regenerator flue gas. The products of the
936 Ind. Eng. Chem. Res., Vol. 29, No. 6, 1990
Flue Gas Product
Catalyst
Maximal 4
Particle
Size, mm
7egeneration Cracking
Regenerated
primary combustions are CO and C02 and HzO, and the Figure 13. Schematic diagram of continuous activity reaction sys-
tem (FCC,TCC).
C02/C0 ratio is a function of temperature. CO reacts with
O2to form C02in a thermal free-radical reaction, but this
reaction is slowed by the presence of solids. If the flue gas 10. Development of the Cracking Reactor
contains excess air, it will ignite as soon as the solids are Having outlined the chemistry of cracking reactions and
removed. Until recently, the design and operation of re- the kinetic constraints, let us now examine how different
generators was dominated by this feature. Nowadays, designs dealt with the challenges. We want to show why
additives containing precious metals, such as platinum, modern FCC technology won over the other designs and
allow complete combustion to C02. examine the chemical reaction engineering lessions that
The combustion of CO to C02 can also be achieved by can be learned for similar design problems.
increasing combustion temperatures. As the activation Three different types of reactors have been developed
energy of homogeneous combustion of CO to C02 is high for catalytic cracking. The fixed-bed reactor enjoyed
(approximately 70 kcal/mol), temperatures of above 700 prominence from 1936 to the early 1940s but is no longer
"C will result in complete CO combustion, given sufficient used today. Only two reactor types are still in use, and
air, but material and catalyst stability constraints reduce only one, the fluidized bed catalytic cracker (FCC), is used
the maximum temperature below that obtainable from the in new designs. Few operating moving bed reactors (TCCs)
heat balance. Catalysts deactivate at high temperatures are left today, and they are being gradually replaced by
by sintering and by steaming. Today's FCC catalysts can fluid-bed crackers. But all three designs have had a lasting
be safely regenerated at temperatures of up to 850 "C impact on chemical reactor design technology.
without excessive damage to them by sintering, but To understood the underlying reasons for reactor choice,
deactivation by steam requires significantly lower tem- let us examine the criteria resulting from the kinetic
peratures. The designer therefore faces a compromise constraints discussed in the previous sections.
between minimizing steam deactivation by lowering tem- (1) Catalytic Cracking Using a Catalyst That
perature and reducing the size of the regenerator by in- Deactivates Quickly Due to Coking. In a fixed-bed
creasing it. Another key to reduce deactivation of steam reactor, each new element of feed meets a catalyst with
is to use two-stage regeneration: the first stage, where lower activity than the previous element of feed. As con-
H2-richcoke is burned, is operated at low temperature; the version depends on the product of catalyst activity and
second stage is operated at a high temperature. space time, this is similar to a reactor with constant cat-
The required residence times for combustion given in alyst activity but nonuniform contact time. This lowers
Figure 11were calculated for an isothermal vessel. FCC conversion and the yield of intermediates in a system of
regenerators are not always isothermal, and one potential consecutive reactions.
source for temperature nonuniformity is mass-transfer One way to overcome changing the catalyst activity is
limitations inside catalyst particles. The impact of mass to move the catalyst continuously through separate reac-
transfer has to be minimized so that the temperature inside tion and regeneration zones (Figure 13). The average
the catalyst particles will not exceed gas-phase tempera- reactivity of the catalyst bed is kept constant by contin-
ture. Diffusion control in combustion is a result of fast uously removing coked catalysts from the bed and adding
reaction rates. Diffusion times vary with the square of the regenerated catalyst to it. Each catalyst particle still ex-
diameter of a catalyst particle, whereas reaction rates are periences a reaction-regeneration cycle, but the feed sees
a function of temperature. constant average catalyst properties.
Figure 12 shows the relation between the maximum Two approaches to realize this concept of moving cat-
particle size and temperature. Particle sizes need to be alyst between reactor and regenerator were developed
below 200 Fm so that the regenerator can operate above simultaneously around 1940. Both used the conceptual
650 OC while avoiding high intraparticle temperatures. scheme illustrated in Figure 13, and both are still being
The minimal particle size in fixed bed reactors is 1 mm used for many other applications today. One used a
whereas in the moving bed it is about 3 mm. Only fluid moving bed with large catalyst particles; the other used
beds operate with particles below 200 Fm. In processes fluidized particle technology, which can either be a dense
that have much lower coke deposition per unit of feed, fluidized bed or an entrained bed.
there are other ways to protect the catalyst. A common (2) Effective Heat Transfer and Utilization. The
way is to use dilluted combustion air. This reduces the cracking reactor must be able to transfer heat to the re-
adiabatic temperature rise of the air-carbon reaction but action at temperatures of 500-550 OC. Cracking generates
is impractical in FCC due to the large volumes of air in- large amounts of coke, and the heat of combustion must
volved. be removed in a controllable way while utilizing the heat
Ind. Emg. Chem. Res., Vol. 29, No. 6, 1990 937
of combustion efficiency. In principle, one can use most PRODUCT
FCC unit is heat integrated on the reactor side as well, and CHARGE
the feed in preheated by heat exchange with products. STOCK
PREHEATER
Other heat-exchange means, such as fired feed preheater,
and an external cooler to remove heat from the regenerator FLOW SHEET, HOUNDRY PROCESS
are also used to provide flexibility and controllability. Figure 14. Schematic diagram of the fixed-bed catalytic cracking
The regenerator is not always in balance with the heat unit.
requirements of the reactor. It is therefore desirable to A~xAIR AND OIL VAPOR OUTLET
design the system so that heat can be both added and
removed independently. The regenerator itself should be I
Tm
local flow rate, and local reaction rates. This puts strong
limitations on the maximum allowable local reaction rates. MAIN TUBE SHEET
The requirements of efficient heat removal and heat ad-
lm
dition are often the dominating parameters in the choice
SALT OUTLETS
of the reactor.
The heat-transfer requirement, together with the re-
quirement of catalyst reactivation, dominated the devel-
opment of hardware and catalyst technologies. The de-
signs that were developed in the past 50 years form the
basis for many other present reactor designs. VAPOR TUBES-
(55)
(3) Operating at High Temperatures. Operation at LEGEND
a
higher temperatures only became desirable for high-ac-
tivity catalysts that were not available at the time of
AIR OR VAPOR PATHS
SALT PATHS
0CATALYSTSPACE
L L SOIL
AIR AND A LVAPOR
T i N INLET
L E T
different designs developed. Short residence times are NOTE DRAWING IS NOT TO SCALE DISTANCES ARE APPROXIMATE ONLY A FEW TUBES ARE
SHOWN FOR THE PURPOSE OF FOLLOWING THE OIL FLOW SMALL ARROWS INDICATE
possible in fixed beds, but short residence times at high PATH OF AIR AND OIL VAPORS THRU THE CASE
temperatures require very rapid heating of the feed. If the Figure 15. Diagramatic sketch of commercial Houdry case.
heating of the feed is done by hot catalyst particles, then
the temperature of the particles is significantly higher than fixed bed (Figure 14). It was jointly developed by Eugene
the averge cracking temperature. Mixing must be rapid Houdry, Sun Oil, and Socony Vacuum, the forerunner of
to avoid thermal cracking. This strongly limits the particle Mobil. The first 2000-BPD unit was started up in April
size, as the heating time of 3-mm particles is about 2 s. 1936 in the Paulsboro, NJ, refinery. Its reactors contained
(4) Operating with Changing Feed and Changing acid-treated natural clay catalyst. Three reactors were
Product Requirements. Cracking feedstocks can vary operated in parallel, and they were switched from cracking
widely, and the cracker must be able to handle a wide to regeneration every few minutes. The heat transmission
variety of feedstocks. Product demands vary seasonally. between cracking and regeneration was achieved by a
Modern fluidized bed crackers are able to meet these de- molten salt heat exchanger. The molten salt was flowing
mands by changing catalyst type or activity, reaction through the reactor in a series of tubes that were equipped
temperature, and space velocity. with fins to increase heat transfer (Figures 15 and 16).
Changing feed properties and conversion also affects the The molten salt extracted the heat from the reactor op-
heat balance, as coke make and heat requirements of the erating in the combustion mode and transferred it to the
reaction change. In the next sections we will discuss how reactor operating in the cracking mode.
different designs dealt with these issues. The fixed-bed reactor was an ingenious solution to meet
the requirements of a cracking reactor with the technology
11. Fixed-Bed Reactor available then. It represented a major milestone in process
The first commercial catalytic cracking reactor was a technology, including such highlights as the first wide-
938 Ind. Eng. Chem. Res., Vol. 29, No. 6, 1990
TUBE SIZES AIR
1. REGENERATIONTUBES - OUTSIDETUBE - Scm.
INSIDETUBE - 3cm.
2. DISTRIBUTORTUBES - OUTSIDETUBE - 6cm.
INSIDETUBE - 3cm.
TO REACTOR
3. COOLINGTUBES - OUTSIDETUBE - 6cm.
INSIDETUBE - 3cm. /
STEAM I
4. COOLINGTUBES - 3.3cm.
HEATCONDUCTING FN
I S- TUBE
TlON
I ELEVATOR
VESSEL
BLOWER
THERMOCOUPLE 1 AND OIL VAPORS
ARRANGEMENT t ENTER CATALYST BE13 Figure 18. Schematic diagram of the TCC air lift process.
u
(AIR AND OIL VAPORS LEAVE CATALYST BED
The other requirements listed in section 10, namely
THROUGH COLLECTOR TUBE ORIFICES)
operating at high temperatures and short residence times,
NOTE DRAWING IS NOT TO SCALE and the ability to handle a large variety of feeds, were not
Figure 16. Cross-section sketch of typical tube unit layout for as important in the late 1930s. The two competing
commercial Houdry case. cracking technologies were developed to improve on item
1, the ability to handle a fast-changing catalyst.
Instantaneous Yield Today's salt-bath tubular reactors are descendants of
50 this technology, and in practically all of them, the catalyst
Gasoline
Yield, r-11 1
45
is contained in tubes. Most of them operate with long
times between regeneration (2 weeks to a year). There is
no need to utilize the heat evolved in the regeneration.
Only one major process still operates with very frequent
I-
40 \\ I switching between reaction and regeneration-catalytic
35t
30
\\ I Amorphous Catalyst
I I I \\ I
dehydrogenation of butane and propane (Catofin and
Cadadiene). This reactor is a direct descendant of the
Houdry reactor, but also has moving bed and fluid-bed
competitors. The moving bed process, UOP's Oleflex, has
60 70 80 90 100
been introduced recently. It is based on the moving bed
Figure 17. Integrated gasoline yield in a fixed-bed reactor vs in- CCR reformer, which in turn is based on TCC principles.
stantaneous yield achievable in a constant activity reactor (based on Fluid-bed technology has been practiced in the Soviet
FCC kinetic model (Jacob et al., 1976)). Union and Romania.
spread use of catalysis in petroleum refining, air regen-
eration of coked catalysts, and the implementation of au- 12. Moving Bed Reactor (TCC)
tomatic process control. however, considering the demands A schematic representation of a TCC unit is given in
of the cracking reactor given in section 10, the fixed-bed Figure 18. The moving bed reactor uses large catalyst
reactor had shortcomings. particles (3-5-mm spheres). The catalyst moves by gravity
(1)Fractions of feed saw an increasingly deactivated from the reactor section to the regenerator section, and
catalyst. The first fraction was overcracked, whereas the the regenerated hot catalyst is lifted by an airlift to a
last fraction had low conversion. This is similar to a re- solids-air separator from which the catalyst is fed by
actor with a broad residence time distribution. Conversion gravity to the reactor. A small side stream is fed to an
is integrated over different histories, which reduces in- elutriator to remove fines generated by attrition.
termediate yields in consecutive reaction schemes. This The TCC had many advantages over the fixed-bed re-
is illustrated in Figure 17, where the integrated yield is actor and allowed much larger units. It solved one of the
compared to the instantaneous yield. major problems of the Houdry reactor, namely, constantly
If catalyst deactivation is slow, one can compensate for changing catalyst activity. The mechanical problems of
it by raising the temperature. This is impractical in cat- such a system were formidable but were solved by elegant
alytic cracking, as deactivation is too fast and the tem- design. This technology has found important new uses
perature is already high. The large thermal holdup of the such as in UOP's continuous catalytic reformer (CCR) and
salt bath would not allow such a control scheme. This Oleflex processes.
would be possible if the time scale of the deactivation was While the TCC had large advantages over the fixed-bed
measured in weeks instead of minutes. reactor, it still fell short in meeting the requirements given
(2) Temperature distribution was not uniform during in section 10. The large particle size limited the regen-
combustion, thus limiting the maximal regeneration tem- erator temperature to below 650 "C, requiring a large re-
perature. generator and large catalyst holdup (Prater et al., 1983).
Ind. Eng. Chem. Res., Vol. 29, No.6, 1990 939
1.o 1.o
lh Bubbling 0.1
Turbulent 0.5
Fast 3 -
-
.
0.1
Riser 10 - Solid
Volume 0.01
Height 0.5
Fraction
Non-Dimentional
Turbulent
0.001
Bubbling
0 I \ I I -___
n.ow1
0 0.2 0.4 0.6 0.8 0.001 0.01 0.1 10 10 100
Solid Volume Fraction Superficial Gas Velocity ( mis)
Figure 19. Axial density profiles in fluid beds. Figure 20. Modern high-efficiency FCCU with riser reactor and
riser regenerator-fluidization regimes (from Squires et al. (1985)).
(Modern TCC catalysts allow regenerator temperatures of FLUE GAS
up to 675 "C, in properly designed kilns.) Furthermore, VENT REGENERATED
design (1942), and the 1952 vintage Model IV is shown in efficiency”regenerator shown in Figure 20. The turbulent
Figure 22. Model I was an “upflow” system, while Model fluid bed at the bottom of the combustor is turned into
IV was a “pressure-balanced downflow” system. Model I, an entrained fast fluid bed by reducing the vessel diameter.
which was the first commercial fluid-bed system using a The solids are separated from the gas in an inertial sepa-
fine powder, had many advanced features that were rator in the upper combustor. Part of the regenerated
dropped in later models but eventually were reintroduced. catalyst is returned to the base of the riser reactor, and
It had upflow cracking and regeneration, quick separation part is recirculated to the lower combustor. This design
of cracked products from spent catalyst, and catalyst resulted in a smaller vessel and better gas-solids contact
coolers and feed preheat (”non-heat-danced” operation). inherent in high-velocity fluidization regimes. The high-
Model I and most later FCC designs (except for Model IV) velocity regenerator is more forgiving toward initial dis-
share some basic features: Catalyst flow rates can be ad- tribution problenis and allows reduction in unit inventory.
justed over a wide range, allowing for great flexibility in It also allows complete CO combustion with little excess
opertion. air and little or no promoter.
In this paper we are not concerned with a detailed Another new regenerator type uses two-stage combus-
discussion of the design changes that led to todays different tion. The fast-burning hydrogen-rich compounds are
FCC units, but rather in the way the demands outlined combusted first at lower temperature in the first regen-
in section 10 effected the design. Let us start with the erator, and the slow-burning residual coke is combusted
regenerator at a high temperature in the second regenerator. This
concept has been commercialized in some resid cracking
14. FCC Regenerators FCCU’s. The goal of this design is to reduce the exposure
Early FCC regenerators operated at tempertures below of the catalyst to steam generated in the combustion, as
650 “C. At these temperatures, there was little CO com- the hydrogen content of the coke and therefore the steam
bustion in the gas phase. The flue gas ignited once the content in the flue gas are lower in the high-temperature
solids were removed from the gas in the cyclone due to the zone. The two-stage design reduces catalyst makeup and
presence of CO and oxygen, and the absence of solids that increases the flexibility of the unit.
quench free radicals. This afterburning could lead to ex-
cessively high temperatures in the cyclones, and oxygen 15. FCC Reactor
conversion needs to be high, leaving less than 0.2% oxygen The Model I FCC reactor was an upflow pipe or riser,
in the flue gas. Such high conversions are difficult to with an expanded diameter section designed to provide
achieve in low-velocity fluid beds without a catalytic more residence time. The upflow mode is also used today
combustion promoter (if low carbon level on regenerated but at much shorter residence times. The designers of this
catalyst is desired). There is significant bypassing in a early fluid-bed system soon discovered that fine powders
large bubbling bed, especially in shallow regenerator beds do not seem to obey Stokes’ law. Even at velocities several
that were designed to minimize pressure drop and air times higher the single-particle terminal velocity (which
blower costs. The practice of operating with low fines is approximately 0.1 m/s for FCC catalyst), the bed is not
concentrations (because of low efficiency cyclones) caused entrained immediately. The particles behave as clusters
large bubbles and aggravated the problem. with an effectively higher terminal velocity due to inter-
Fluid-bed regenerator efficiency can be improved by particle forces. This realization led to the development
operating at higher velocities such that there is a significant of Model I1 and subsequent “downflow”fluid-bed reactors.
catalyst concentration in the freeboard. A typical FCC The expanded reactor section was elongated, a dense fluid
regenerator operates in turbulent fluidization at a gas bed was formed, and the solids were withdrawn from the
velocity of 1 m/s, such that about 20% of the catalyst is bottoms, rather the top of the bed. The higher bed density
in the freeboard. In fact, whole bed circulates through the in these downflow system allowed long contact times and
cyclones about every 5 min. This solids holdup in the lowered catalyst losses which were high in Model I.
freeboard allows relatively high conversions in these Improvements in catalysts, brought about by introduc-
shallow, inefficient regenerators. However, afterburning tion of synthetic silica-alumina to replace natural clays,
in the cyclones can still be a problem. The temperature followed by high alumina catalysts, caused more of the
rise due to afterburning is used to control the regenerator. cracking reaction to occur in the riser leading to the re-
To achieve higher CO conversion, one needs to provide actor. Mobil’s invention of zeolite-containing cracking
combustion to COz either by raising the temperature catalysts in the early 1960s has eliminated the need for the
(above 700 “C) or by utilizing a catalytic combustion dense bed altogether. Existing FCC reactors were modified
promoter. Small amounts of Pt allow complete CO con- to all riser cracking, usually by extending the riser through
version at temperatures of 650 “C. The activation energy the reactor vessel. The “reactor”vessel, which is still called
of CO combustion with a catalyst is only 20 kcal/mol reactor in many installations, now houses the riser cyclones
(Prater et al., 1983). and, in some designs, the catalyst stripper (which uses
Even early FCC regenerators benefited from nearly steam to remove some adsorbed hydrocarbon from spent
isothermal conditions. Hot spots, which can easily develop catalyst).
in fixed-bed or moving bed reactors (there is no lateral The modern FCC reactor is different from conventional
mixing in TCC), are eliminated in fluidized beds. For the catalytic fluid-bed reactors. The catalyst is introduced
same reason, fluid-bed chemical reactors (such as acrylo- through a standpipe, and oil feed is introduced into the
nitrile) can be operated in the explosive regime, without bottom of the riser section. This allows rapid mixing be-
ever exploding. Still, fluid bed reactors are not completely tween the vaporizing feed and hot regenerated catalyst.
isothermal. Lateral and vertical temperature differences The bottom of the riser is a complex environment. There
can reach 50 “C, and larger temperature fluctuations can are strong axial and radial temperatures and solid con-
occur over very short time periods. But the impact is much centration profiles. The desired goals for the mixing zone
less than that of excess temperatures inside a large catalyst are rapid mixing of solids and oil and minimizing catalyst
particle. holdup in the zone. A high catalyst holdup indicates that
FCC regenerators underwent many changes over the vaporization is slow and that, therefore, the oil residence
years. One particularly successful design is the “high- time is larger than indicated by average gas velocity.
Ind. Eng. Chem. Res., Vol. 29, No. 6, 1990 941
b) Time-Temperature Regimes for Fluid-Solids Systems Major Control Loops
(Other Loops Omitted for Clarity)
Regenerator Reactor
2000 r QC Reaction UJ I-- To Main
Fractionator
A-P
+
(P+ I D)
Set
Temp.,lmE
‘F
Art
500 I I I I .
0.1 1.0 10 100
Generation
Residence Time, Sec.
1
SolldS
Air
Furnace
Figure 23. QC reaction system (from Gartside (1989)).
Oil Fuel
Reactor design in modern FCC units still falls short of
the goals outlined in section 10. The achievable residence Figure 24. One FCC control scheme with steam generation in the
regenerator (Shinnar, 1981).
time is still larger than desired to achieve the optimum
ratio of chemical to catalytic cracking at high temperatures.
The limitations are in the gas-solids separation as well as to summer. It also allows trade-offs between gasoline yield
in the bottom mixing zone. Recent advances in quick and octane. The fluid-bed unit has a greater degree of
gas-solids separation have moved the main constraint into flexibility than both the fixed-bed and moving bed reac-
the bottom mixing zone. Vaporization and cracking re- tors. The degree of flexibility is a strong function of design
quire milliseconds for an active catalyst, but it is difficult and varies strongly in different designs. The critical items
to mix large amounts of catalyst (over 60 tons/min in a which allow FCC to operate over a wide range of conditions
large unit) with cold oil feed within milliseconds. There are as follows.
is therefore a strong variation in the temperature-time (a) Catalyst. Continuous (or daily) catalyst addition
activity history that different elements of the feed undergo. and withdrawal of equilibrium catalyst compensate for
In that sense, the effective contact time distribution is permanent deactivation. The catalyst change in FCC is
highly nonuniform. One can measure the impact of this much more rapid than in TCC due to the lower inventory.
nonuniformity by comparing the overall results with that The whole inventory in FCC is typically replaced every
of a small pilot plant. The differences are important 50-100 days, and response to a new catalyst or additive
enough to merit improvement, but they are not large is much quicker.
enough to outweigh the other advantges of this design. (b) Reactor Temperature. Temperture is an inde-
Faster mixing and shorter residence times are still an pendent variable that adjusts steady-state catalyst activity
area where there is still a place for novel designs. One and allows changing product slate and properties (such as
interesting development is the “quick contact” (QC) system gasoline octane).
described in Figure 23. This sytem can contact feed and (c) Catalyst Circulation Rate. Catalyst circulation
catalyst, react them, and separate and quench them in less rate can be adjusted over a wide range in most FCC de-
than half a second. (Current systems need more than 1 signs. In a truly adiabatic FCC, in which there is no feed
s.) However, no commercial plants are yet in operation. preheat and no catalyst cooler, the circulation rate for a
Understandingand modeling highly nonisothermal mixing given reaction temperature is mainly a function of regen-
systems in which residence times and mixing times are of eration temperature and therefore strongly constrained.
comparable magnitude is still a challenge for the reaction If a preheat furnace is added to the unit, the catalyst
engineer. circulation rate can be adjusted over a much wider range.
There is another interesting lesson for the reaction en- Catalyst coolers are also used to adjust FCC heat balance.
gineer. The conditions at the bottom zone of a modern Catalyst type also affects circulation rate. A coke-selective,
FCC riser are far beyond our capabilities to realize detailed ultrastable catalyst would not change the coke yield in a
scaleup by predictive modeling. This was less severe a heat-balanced unit but would lower the carbon on spent
problem in earlier designs as catalysts were less active and catalyst. The C/O ratio will hence increase.
temperature differences were much smaller. We have to (d) C 0 2 / C 0 Ratio. Another important factor im-
realize that sometimes successful designs involve risks. pacting the heat balance is the C02/C0 ratio in the re-
Such risks can be minimized but not avoided. generator. If no combustion promoter is used, this ratio
is solely a function of regenerator temperature. Pt pro-
16. Design for Controllability moter allows operation at full CO combustion, at any op-
One important consideration in the choice and design erating temperature in most units. If one desires to operate
of a reactor is the ability to control it (Shinnar, 1981; Lee at partial combustion, a CO boiler is needed to complete
and Weekman 1976). In a reactor that has the potential the combustion of the flue gas. Use of a promoter allows
for multiple steady states, it is also important that the fine tuning of this ratio and thereby control of the heat
dynamic control be able to maintain unconditional as- balance.
ymptotic stability. In addition, the cracker has to handle These and other control items are combined into an
a wide variety of feedstocks and to change the product overall control scheme, such as the one shown in Figure
distributions and product specification, upon short notice. 24 (Shinnar, 1981). The two main variables that would
FCC’s ability to accept a wide range of feedstocks makes affect the conversion are reactor riser top temperature,
it the kingpin in controlling the whole refinery. Changing which controls the catalyst circulation rate, and feed
FCC operating conditions allows changes in the ratio preheat temperature. One important control variable,
gasoline to heating oil, when demands change from winter available in the first Model I FCC unit and recently res-
942 Ind. Eng. Chem. Res., Vol. 29,No. 6,1990
urrected, is heat removal from the regenerator. Steam is n = Voorhies exponent, about 0.5 in Voorhies’ work
generated in coils or an external catalyst cooler. This Ri = rate matrix for first-order reaction network (eq 15)
arrangement is especially useful when processing heavy t , = catalyst residence time, min
feeds, such as resids with high CCR levels. TCC = thermofor catalytic cracking
Modern FCC design meets most of the criteria outlined WHSV = weight hourly space velocity
in section 10 well and definitely meets most of those X = conversion (=lo0- vol % yield of products heavier than
criteria better than any of the other designs. There are gasoline, typically above 220 OC)
still two areas in which FCC designs fall short. X = Crackability, (=X/(1- X))
(1)The lower limit of residence time is still higher than AG = change in Gibbs free energy
desirable, mainly due to limitation of the bottom section AH = change in enthalpy
@ = catalyst deactivation function
and the gas catalyst separation. A, = eigenvalues of the rate matrix Ri
(2) The contact time distribution is not uniform. Fur- T = time constant related to WHSV in eq 16
thermore, the temperature activity history of different fluid ‘;k(t)= contact time distribution
elements can vary considerably. *(t) = Laplace transform of Jr(t)
These shortcomings still leave room for future imagi-
native design. But any design is a compromise, and the Literature Cited
present designs are better compromises than former Avidan, A. A.; Edwards, M.; Owen H. Innovative Improvements
technologies. We hope that the previous discussion is Highlight FCC’s Past and Future. Oil Gas J. 1990,88 (2), 33-58.
useful for the student in reaction engineering in teaching Blanding, F. H. Znd. Eng. Chem. 1953,45 (6), 1186-1197.
him that reactor design has to take into account a large Chen, N. Y.; Lucki, S. J. Ind. Eng. Chem. Process Des. Deu. 1986,
variety of factors and always involves compromises. 25 (3), 814-820.
Gartaide, R. J. QC-A New Reaction System In Fluidization VI;
Acknowledgment Grace, J. R., Schemilt, L. W., Bergougnou, M. A., Eds.; Engi-
neering Foundation: New York, 1989.
The figures in this paper are intended to demonstrate Jacob, S. M.; Gross, B.; Voltz, S. E.; Weekman, V. W., Jr., AZChE
trends rather than show exact data. We are grateful to J . 1976,22, 701-713.
Lee, W.; Weekman, V. W., Jr. AIChE J. 1976,22,27.
Raymond Hu for his assistance with Figures 1 , 2 , 7 , 8 , 11, Prater, C. D.; Wei, J.; Weekman, V. W., Jr.; Gross, B. In Advances
12, and 16. Lee Breckenridge supplied us with Figure 5. in Chemical Engineering; Wei, J., Ed.; Academic Press: New
Figures 4 and 6 are drawn after several authors, including York, 1983; Vol. 12, pp 1-60.
Voorhies; orders of magnitude was taken from Mobil re- Sapre, A. V.; Leib, T. M. Translation of Laboratory FCC Catalyst
searchers Farber, Payne, and Sailor (1965). Most other Characterization Tests t o Riser Reactors; Advances in FCC
figures are adapted from Avidan et al. (1990). Technology; American Chemical Society: Washington, DC, Aug
1990.
Shinnar, R. Chem. Eng. Commun. 1981,9,73-99.
Nomenclature Shinnar, R. Use of Residence and Contact Time Distributions in
A = constant in Voorhies equation (0.2-0.9 in ref) Reactor Design. In Chemical Reactor and Reaction Engineering;
Aj, = constant determined by feed Carbury, J., Varma, A,, Eds.; Marcel Dekker: New York, 1986.
Shinnar, R.; Glasser, D.; Katz, S . Ind. Eng. Chem. Fundam. 1973,
C = coke yield, wt % of feed 12, 165.
C, = carbon deposited on catalyst, wt 70 (carbon on spent Squires, A. M.; Kwauk, M.; Avidan, A. A. Science 1985,230 (Dec 20),
carbon on regenerated catalyst) 4732, 1329-1337.
CCR = Conradson carbon residue, or continuous catalyst Voorhies, A., Jr. Znd. Eng. Chem. 1945,37 (4), 318-322.
regeneration (when applied to reforming) Weekman, V. W., Jr. Ind. Eng. Chem. Process Des. Deu. 1969,B (31,
Cj= outlet reactor composition (eq 16) 385-391.
C/O = catalyst to oil weight ratio Weekman, V. W., Jr.; Nace, D. M. AIChE J. 1970,16 (3), 397-404.
FCC = fluid catalytic cracking Weisz, P. B.; Goodwin, R. D. J. Catal. 1963,2 (5), 397-403.
H/C = hydrogen to carbon atomic ratio in coke Wojciechowski, B. W.; Corma, A. Catalytic Cracking;Marcel Dek-
ker: New York, 1986.
I&EC = Industrial & Engineering Chemistry
k, = coke yield divided by crackability, X Received for review October 6, 1989
kj = catalyst activity for each reaction shown in eq 8 Revised manuscript received January 18, 1990
k,, kj, = catalyst activity for regenerated catalyst Accepted February 22, 1990